Ujex
Ujex is open-source agent trust infrastructure. It gives AI agents the pieces production teams need before they let software act on their behalf: signed identity, approval queues, provenance logs, policy guardrails, secrets, tool authorization, and evidence exports.
ujex identity passport agent-007 > passport.json
ujex send billing@vendor.com "renewal" "Approved terms attached." --session=sess_renewal_001 --require-human
# Export the signed evidence bundle from app.ujex.dev Evidence.
ujex audit verify ./evidence
The golden path is simple:
- Create an agent and fetch its signed Agent Passport.
- Run one approval-gated action.
- Export a signed evidence bundle that another person can verify offline.
:::tip What makes Ujex different? Most agent frameworks are code libraries. Ujex is a trust stack: Cloud Functions, Security Rules, storage, scheduled jobs, SDKs, CLI, and docs ship together so identity, approvals, policy, and audit evidence stay linked. :::
Why
Agents can already reason, write, and call tools. The missing production layer is accountability:
- Who is this agent? A signed passport, not a screenshot in a dashboard.
- Was the action allowed? Policy, risk, and approval decisions before work runs.
- What credentials were used? Server-injected tool credentials, scoped runtime secrets, and audit rows.
- Can we prove what happened? Hash-chained records and signed export manifests.
Ujex focuses on those trust primitives. It is not a marketplace, checkout system, escrow service, browser fleet, GPU pool, recruiting tool, robotics stack, or agency.
Core stack
| Primitive | What it proves | Status |
|---|---|---|
| Identity | Agent Passport, scoped device keys, delegation, owner hash | Live |
| Inbox | Agent inbox, prompt-injection risk signals, approval-gated sends | Core · Live |
| Approvals | Human decision, linked action digest, expiry, audit record | Core · Live |
| Audit / Evidence | Hash chain, AAT export, signed manifest, anchors | Core · Live |
| Budgets / Guardrails | Estimated action-based spend and risk decision | Core · Live |
| Tools | Scoped tool manifests and server-injected credentials | Core · Live |
| Provenance | Searchable markdown memory with audit links | Secondary · Live |
| Advanced compatibility | Useful infrastructure, kept out of the core buyer path | Hidden from first-run |
What's deployed
Reference deployment runs on Firebase project axy-ujex.
| Surface | Domain | Build target |
|---|---|---|
| Public site | ujex.dev, compliance.ujex.dev | Vite landing |
| Docs | docs.ujex.dev | Docusaurus |
| App dashboard | app.ujex.dev | Vite app |
| Admin console | admin.ujex.dev | Vite admin |
| Control plane | Cloud Functions in us-central1 | axy-ujex |
| Mail bridge | mail.ujex.dev | Postfix/OpenDKIM relay |
Where to go next
- First time here? Start with the Quickstart.
- Need evidence? Read Evidence verification.
- Evaluating risk? Review Threat Model.
- Integrating? Use the API Reference or SDK guide.
:::info Last verified 2026-06-21 · public docs aligned to the agent trust infrastructure wedge. :::